【作者】 毛政利；

【导师】 彭省临；

【作者基本信息】 中南大学 ， 矿产普查与勘探， 2004， 博士

【摘要】 铜陵地区是我国铜矿较集中发育的地区之一，位于长江中下游铜铁金多金属成矿带的中部。从该区及其邻区的沉积建造、岩浆建造、变质建造、构造型相及矿床分布的时空规律等分析，其大地构造演化经历了前地槽、地槽、地台、地洼四个阶段，不同的演化阶段其成矿作用不同，其中以地洼阶段的燕山期岩浆活动为主的成矿作用最为强烈，本区大部分矿床在这一时期的岩浆活动中基本定型。 凤凰山铜矿是铜陵矿集区主要矿床之一，它赋存于新屋里岩体西部的接触带上，矿体的产出受新屋里岩体、三叠系灰岩、构造、特别是与岩体同期形成的脆性扩容断裂构造的复合控制。脆性扩容断裂构造带中普遍发育有角砾岩，包括角砾状花岗闪长岩、角砾状矿石和角砾状大理岩等三种主要类型。经分形研究，它们的形态和大小分维值分别为：1.26，1.14，1.01和1.389，1.526，1.24，说明角砾状花岗闪长岩主要是化学成因形成的，而角砾状矿石和角砾状大理岩主要是由机械作用的水力致裂形成的，形成时的能量角砾状矿石大于角砾状大理岩。 凤凰山铜矿床南区构造地球化学取样分析中的Cu元素含量具有多重分形的特点，且不同的地质体中其分形特征不同，结合Cu元素在不同地质体中的分布特点，说明本区不同地质体所经历的成矿作用不同，其中以石英二长闪长岩所经历的成矿作用强度较大，较有利于成矿元素的富集。 BP神经网络技术是目前常用的一种非线性拟合方法，它的作用在于提供了一种非线性静态映射，能以任意精度逼近任意给定的非线性关系，具有学习和推广的功能。它通过各神经元之间的藕合作用关系实现从输入到输出的高度非线性映射。矿体的形成与定位也是各种控矿因素祸合作用的结果，其祸合机制是线性函数难以拟合的，而BP人工神经网络的高度非线性拟合功能恰好能解决这一问题。利用凤凰山铜矿床8个已知样本，将其输入网络学习，经过5O0()余次迭代，网络收敛，其输出误差满足精度要求。因此，利用BP人工神经网络模型来拟合成矿作用过程中各种控矿因素之间的藕合作用、并以此建立木区的大比例尺成矿预测模型是可行的。即人工神经网络的非线性拟合功能主要是通过网络内部各神经元之间的连接权值实现的，因而神经元的连接权值定量地表达了该神经元对下一层神经元的影响权重，通过网络运行机理和对神经元之间的连接权值的追踪，可计算出输入层神经元对输出层神经元的影响权重，结合地球物理、地球化学勘查数据，建立了基于成矿有利度法的凤凰山铜矿床找矿预测模型。在组件式GIS软件MapX平台上，利用可视化高级编程语言vi Sualc+十将成矿预测模型、找矿预测模型和地学空间数据库集成在一起，开发实现了凤凰山铜矿床成矿预测系统，完成本矿床的大比例尺数字矿床模型。更多还原

【Abstract】 The Tongling region is one of the region that copper ore deposits were concentrated developed in our country. It locate in the middle of the Cu-Fe-Au multimetal mineralization belt of middle and lower reaches of Yangtze River. Its geotectonic evolution has gone through four stages: Pre-geosynclinal stage, Geosynclinal stage, Platform stage and Diwa stage by the analyzing of sedimentary formation, magma formation, metamorphic formation, structural styles & tectofacies and space-time orderliness of mineral distribution in this region and its vicinage. Every stage had its different metallogenesis. Among these, the strongest metallogenesis is the one which arose by Yanshanian magmatic processes in Diwa stage. The most ore deposits were formed during this magmatic processes in the region.Fenghuangshan copper deposit is one of the mainly deposit in Tongling ore deposits concentrated region. It lodged in the western contact strip of Xinwuli rock body. The occurrence of ore body was controlled by Xinwuli rock body, Triassic limestone, structure, especial the fragile dilatant fault structure which was formed in the same period of time with rock body. In the fragile dilatant fault structure belt, there are breccias generally which mainly include three types: brecciated granodiorite, brecciated ore and brecciated marble. According to thefractal analysis, its’ fractal dimension of morphology and size distribution is 1.26,1.14,1.01 and 1.389,1.526,1.24. It indicated that, brecciated granodiorite was foemed mainly by chemical attack, brecciated ore and brecciated marble were hydraulic breccias which were formed mainly by mechanical process, and the energy when brecciated ore was formed is higher than the energy when brecciated marble was formed.During the research of tectono-geochemistry in the south part of Fenghuangshan copper deposit, it is discovered that the fractal character of Cu is multifractality, and different from each other according to the geological bodies. Compound with the distribution character of Cu in different geological bodies, it is illustrated that the metallogenesis which different geological bodies had gone through in this area is different, and the intensity of metallogenesis which quartzite monzodiorite had gone through is higher, and in favor of the concentration of metallogenic element.BP neural network technique is a in common used non-linear fit method now. Its function is that it provide a non-linear static mapping, and can gain upon a non-linear relation which discretionary imparted with a discretionary precision. It also has the learning and extending function. It can achieve the highly non-linear mapping from input to output by the coupling relations among neural cells. The formation and location of ore body is the output of the coupling of every factor of ore controlling. Itscoupling mechanism is difficult to fit with linear function. The highly non-linear fit function of BP artificial neural network can resolve yhis problem appropriately. Using eight known specimens of Fenhuangshan copper deposit, and putting its itto the network to learn, after more than 5000 iterativenesses, the network constringed and the error of output reached the requiration of precision. There fore, it is feasible that using the model of BP artificial neural network to fit the coupling among the factor of ore controlling during the ore forming process and the model of metallogenic prediction of this area is established by this way. The non-linear fit function of BP artificial neural network is achieved mainly by the connecting weights among the neural cells inside the network. And so, the connecting weight quantificationally show the weight which the neural cell influence next layer’s neural cell. The influence weight from input layer’s neural cell to output layer’s neural cell can calculate by the running mechanism of network and the tracing weight of neural cell. Based on this weights, and compound with the exploring data of geophysics and geochemistry, the model of prospecting prediction of Fenghuang更多还原